Submission ID: 6d887e04
DISK-depth
Processed: 21-06-16. Download link: 6d887e04583a4e9c-disk-depth-2048.json
This page ranks the submission against all others using the same number of keypoints, regardless of descriptor size. Please hover over table headers for descriptions on metrics and full scene names.
- Phototourism dataset: Stereo track / Multiview track
- Prague Parks dataset: Stereo track / Multiview track
- Google Urban dataset: Stereo track / Multiview track
Metadata
- Authors: Michal Tyszkiewicz (contact)
- Keypoint: disk
- Descriptor: disk (128 float32: 512 bytes)
- Number of features: 2048
- Summary: Official DISK submission with 2k features, using ratio test and cycle-consistent matching, with DEGENSAC.
- Paper: https://proceedings.neurips.cc/paper/2020/file/a42a596fc71e17828440030074d15e74-Paper.pdf
- Website: https://github.com/cvlab-epfl/disk
- Processing date: 21-06-16
Phototourism dataset / Stereo track
mAA at 10 degrees: 0.51219 (±0.00000 over 1 run(s) / ±0.12737 over 9 scenes)
Rank (per category): 43 (of 87)
| Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| BM | 2048.0 | 717.9 | 571.9 | 0.583 Rank: 14/87 |
0.958 Rank: 3/87 |
0.27390 (±0.00000) Rank: 44/87 |
0.41462 (±0.00000) Rank: 43/87 |
| FCS | 2048.0 | 514.8 | 400.1 | 0.420 Rank: 8/87 |
0.893 Rank: 5/87 |
0.57609 (±0.00000) Rank: 41/87 |
0.69302 (±0.00000) Rank: 45/87 |
| LMS | 2048.0 | 430.9 | 326.6 | 0.394 Rank: 51/87 |
0.709 Rank: 8/87 |
0.47243 (±0.00000) Rank: 58/87 |
0.59088 (±0.00000) Rank: 60/87 |
| LB | 2048.0 | 452.6 | 349.9 | 0.472 Rank: 3/87 |
0.782 Rank: 2/87 |
0.48342 (±0.00000) Rank: 44/87 |
0.57690 (±0.00000) Rank: 39/87 |
| MC | 2048.0 | 685.6 | 555.8 | 0.525 Rank: 14/87 |
0.959 Rank: 2/87 |
0.37019 (±0.00000) Rank: 44/87 |
0.53079 (±0.00000) Rank: 9/87 |
| MR | 2048.0 | 471.7 | 392.6 | 0.431 Rank: 26/87 |
0.937 Rank: 8/87 |
0.28033 (±0.00000) Rank: 25/87 |
0.38164 (±0.00000) Rank: 31/87 |
| PSM | 2048.0 | 341.6 | 265.0 | 0.358 Rank: 26/87 |
0.730 Rank: 4/87 |
0.13798 (±0.00000) Rank: 51/87 |
0.25557 (±0.00000) Rank: 51/87 |
| SF | 2048.0 | 474.2 | 366.6 | 0.383 Rank: 19/87 |
0.837 Rank: 5/87 |
0.45254 (±0.00000) Rank: 43/87 |
0.57560 (±0.00000) Rank: 47/87 |
| SPC | 2048.0 | 539.1 | 408.4 | 0.468 Rank: 4/87 |
0.865 Rank: 2/87 |
0.43820 (±0.00000) Rank: 43/87 |
0.59065 (±0.00000) Rank: 43/87 |
| Avg | 2048.0 | 514.3 | 404.1 | 0.448 Rank: 8/87 |
0.852 Rank: 2/87 |
0.38723 (±0.00000) Rank: 42/87 |
0.51219 (±0.00000) Rank: 43/87 |
We show the inliers that survive the robust estimation loop (i.e. RANSAC), or those supplied with the submission if using custom matches, and use the depth estimates to determine whether they are correct. We draw matches above a 5-pixel error threshold in red, and those below are color-coded by their error, from 0 (green) to 5 pixels (yellow). Matches for which we do not have depth estimates are drawn in blue. Please note that the depth maps are estimates and may contain errors.
— British Museum —
— Florence Cathedral Side —
— Lincoln Memorial Statue —
— London Bridge —
— Milan Cathedral —
— Mount Rushmore —
— Piazza San Marco —
— Sagrada Familia —
— Saint Paul's Cathedral —
Phototourism dataset / Multiview track
mAA at 10 degrees: N/A (±0.00000 over 1 run(s) / ±0.09580 over 9 scenes)
Rank (per category): 46 (of 87)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| BM | 2048.0 | 716.44 | 100.00 Rank: 1/87 |
2232.42 Rank: 21/87 |
6.501 Rank: 9/87 |
0.40367 Rank: 43/87 |
0.53592 (±0.00000) Rank: 44/87 |
0.68133 (±0.00000) Rank: 44/87 |
| FCS | 2048.0 | 530.76 | 97.08 Rank: 56/87 |
2605.30 Rank: 34/87 |
5.252 Rank: 6/87 |
0.28176 Rank: 64/87 |
0.70136 (±0.00000) Rank: 57/87 |
0.76637 (±0.00000) Rank: 61/87 |
| LMS | 2048.0 | 461.18 | 99.47 Rank: 16/87 |
2112.31 Rank: 19/87 |
5.587 Rank: 32/87 |
0.33588 Rank: 67/87 |
0.79068 (±0.00000) Rank: 63/87 |
0.85350 (±0.00000) Rank: 63/87 |
| LB | 2048.0 | 534.16 | 97.76 Rank: 55/87 |
2016.21 Rank: 35/87 |
6.050 Rank: 6/87 |
0.46384 Rank: 6/87 |
0.73144 (±0.00000) Rank: 5/87 |
0.80701 (±0.00000) Rank: 32/87 |
| MC | 2048.0 | 672.07 | 100.00 Rank: 1/87 |
2526.58 Rank: 9/87 |
6.122 Rank: 12/87 |
0.39748 Rank: 55/87 |
0.52313 (±0.00000) Rank: 44/87 |
0.67414 (±0.00000) Rank: 44/87 |
| MR | 2048.0 | 463.55 | 94.67 Rank: 34/87 |
2544.15 Rank: 7/87 |
4.909 Rank: 23/87 |
0.47975 Rank: 5/87 |
0.45076 (±0.00000) Rank: 6/87 |
0.57261 (±0.00000) Rank: 8/87 |
| PSM | 2048.0 | 338.90 | 96.28 Rank: 47/87 |
2689.36 Rank: 18/87 |
4.271 Rank: 46/87 |
0.54946 Rank: 55/87 |
0.52113 (±0.00000) Rank: 49/87 |
0.59681 (±0.00000) Rank: 49/87 |
| SF | 2048.0 | 465.99 | 97.07 Rank: 57/87 |
2694.05 Rank: 38/87 |
5.059 Rank: 22/87 |
0.29062 Rank: 39/87 |
0.73667 (±0.00000) Rank: 44/87 |
0.81612 (±0.00000) Rank: 50/87 |
| SPC | 2048.0 | 554.10 | 99.15 Rank: 58/87 |
2402.83 Rank: 24/87 |
5.788 Rank: 10/87 |
0.50305 Rank: 57/87 |
0.70141 (±0.00000) Rank: 50/87 |
0.79840 (±0.00000) Rank: 54/87 |
| Avg | 2048.0 | 526.35 | 97.94 Rank: 50/87 |
2424.80 Rank: 22/87 |
5.504 Rank: 10/87 |
0.41172 Rank: 48/87 |
0.63250 (±0.00000) Rank: 42/87 |
0.72959 (±0.00000) Rank: 46/87 |
In the multi-view track we reconstruct the scene with Structure-from-Motion (Colmap) with small sets of images. We show the results for one bag of 10 images. Keypoints are drawn in blue if they are part of the model, and in red otherwise.
— British Museum —
— Florence Cathedral Side —
— Lincoln Memorial Statue —
— London Bridge —
— Milan Cathedral —
— Mount Rushmore —
— Piazza San Marco —
— Sagrada Familia —
— Saint Paul's Cathedral —
Prague Parks dataset / Stereo track
mAA at 10 degrees: 0.45896 (±0.00000 over 1 run(s) / ±0.05348 over 3 scenes)
Rank (per category): 61 (of 87)
| Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| Lizard | 2048.0 | 239.4 | 194.7 | 0.040 Rank: 17/87 |
0.008 Rank: 60/87 |
0.28266 (±0.00000) Rank: 64/87 |
0.38793 (±0.00000) Rank: 64/87 |
| Pond | 2048.0 | 303.1 | 237.0 | 0.069 Rank: 14/87 |
0.059 Rank: 5/87 |
0.34749 (±0.00000) Rank: 62/87 |
0.47194 (±0.00000) Rank: 62/87 |
| Tree | 2048.0 | 195.7 | 149.5 | 0.045 Rank: 11/87 |
0.019 Rank: 59/87 |
0.40839 (±0.00000) Rank: 61/87 |
0.51700 (±0.00000) Rank: 60/87 |
| Avg | 2048.0 | 246.1 | 193.8 | 0.051 Rank: 13/87 |
0.029 Rank: 10/87 |
0.34618 (±0.00000) Rank: 62/87 |
0.45896 (±0.00000) Rank: 61/87 |
We show the inliers that survive the robust estimation loop (i.e. RANSAC), or those supplied with the submission if using custom matches, and use the depth estimates to determine whether they are correct. We draw matches above a 5-pixel error threshold in red, and those below are color-coded by their error, from 0 (green) to 5 pixels (yellow). Matches for which we do not have depth estimates are drawn in blue. Please note that the depth maps are estimates and may contain errors.
— Lizard —
— Pond —
— Tree —
Prague Parks dataset / Multiview track
mAA at 10 degrees: N/A (±0.00000 over 1 run(s) / ±0.13085 over 3 scenes)
Rank (per category): 60 (of 87)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| Lizard | 2048.0 | 285.82 | 82.97 Rank: 57/87 |
827.50 Rank: 50/87 |
3.400 Rank: 25/87 |
13.90366 Rank: 14/87 |
0.51153 (±0.00000) Rank: 58/87 |
0.60169 (±0.00000) Rank: 58/87 |
| Pond | 2048.0 | 424.35 | 69.00 Rank: 24/87 |
995.71 Rank: 24/87 |
3.106 Rank: 7/87 |
0.42760 Rank: 37/87 |
0.23851 (±0.00000) Rank: 61/87 |
0.28189 (±0.00000) Rank: 54/87 |
| Tree | 2048.0 | 262.37 | 71.87 Rank: 23/87 |
568.81 Rank: 24/87 |
3.099 Rank: 4/87 |
5.79027 Rank: 15/87 |
0.37020 (±0.00000) Rank: 60/87 |
0.42337 (±0.00000) Rank: 55/87 |
| Avg | 2048.0 | 324.18 | 74.61 Rank: 45/87 |
797.34 Rank: 35/87 |
3.202 Rank: 6/87 |
6.70718 Rank: 15/87 |
0.37341 (±0.00000) Rank: 60/87 |
0.43565 (±0.00000) Rank: 60/87 |
In the multi-view track we reconstruct the scene with Structure-from-Motion (Colmap) with small sets of images. We show the results for one bag of 10 images. Keypoints are drawn in blue if they are part of the model, and in red otherwise.
— Lizard —
— Pond —
— Tree —
Google Urban dataset / Stereo track
mAA at 10 degrees: 0.27638 (±0.00000 over 1 run(s) / ±0.13494 over 17 scenes)
Rank (per category): 46 (of 87)
| Scene | Features | Matches (raw) |
Matches (final) |
mAA(5o) | mAA(10o) |
| Amsterdam (AMS) | 2048.0 | 254.5 | 220.9 | 0.15804 (±0.00000) Rank: 42/87 |
0.26295 (±0.00000) Rank: 45/87 |
| Bangkok (BGK) | 2048.0 | 308.8 | 257.3 | 0.06402 (±0.00000) Rank: 19/87 |
0.15732 (±0.00000) Rank: 34/87 |
| Barcelona (BCN) | 2048.0 | 151.1 | 124.3 | 0.09084 (±0.00000) Rank: 43/87 |
0.15348 (±0.00000) Rank: 44/87 |
| Buenos Aires (BAR) | 2048.0 | 190.0 | 159.1 | 0.18721 (±0.00000) Rank: 42/87 |
0.30639 (±0.00000) Rank: 41/87 |
| Cambridge (CAM) | 2048.0 | 217.5 | 185.6 | 0.25822 (±0.00000) Rank: 42/87 |
0.38562 (±0.00000) Rank: 41/87 |
| Cannes (CAN) | 2048.0 | 308.7 | 259.9 | 0.39631 (±0.00000) Rank: 26/87 |
0.49724 (±0.00000) Rank: 32/87 |
| Chicago (CHI) | 2048.0 | 61.7 | 39.4 | 0.04044 (±0.00000) Rank: 48/87 |
0.06776 (±0.00000) Rank: 49/87 |
| Helsinki (HEL) | 2048.0 | 389.1 | 332.4 | 0.33041 (±0.00000) Rank: 35/87 |
0.52912 (±0.00000) Rank: 36/87 |
| Madrid (MAD) | 2048.0 | 130.5 | 101.8 | 0.03465 (±0.00000) Rank: 46/87 |
0.08571 (±0.00000) Rank: 47/87 |
| Mountain View (MTV) | 2048.0 | 116.5 | 90.3 | 0.10859 (±0.00000) Rank: 48/87 |
0.18125 (±0.00000) Rank: 48/87 |
| New Orleans (NOR) | 2048.0 | 140.3 | 109.7 | 0.20414 (±0.00000) Rank: 44/87 |
0.25917 (±0.00000) Rank: 44/87 |
| San Francisco (SF) | 2048.0 | 197.6 | 165.4 | 0.13176 (±0.00000) Rank: 45/87 |
0.22992 (±0.00000) Rank: 46/87 |
| Singapore (SG) | 2048.0 | 205.8 | 165.7 | 0.18608 (±0.00000) Rank: 45/87 |
0.27373 (±0.00000) Rank: 46/87 |
| Sydney (SYD) | 2048.0 | 157.6 | 119.7 | 0.22882 (±0.00000) Rank: 41/87 |
0.35022 (±0.00000) Rank: 42/87 |
| Tokyo (TOK) | 2048.0 | 237.7 | 187.0 | 0.37157 (±0.00000) Rank: 47/87 |
0.50515 (±0.00000) Rank: 47/87 |
| Toronto (TOR) | 2048.0 | 88.6 | 64.3 | 0.10480 (±0.00000) Rank: 47/87 |
0.21120 (±0.00000) Rank: 47/87 |
| Zurich (ZRH) | 2048.0 | 168.5 | 131.7 | 0.19038 (±0.00000) Rank: 46/87 |
0.24227 (±0.00000) Rank: 46/87 |
| Average (Avg) | 2048.0 | 195.6 | 159.7 | 0.18155 (±0.00000) Rank: 45/87 |
0.27638 (±0.00000) Rank: 46/87 |
We show the inliers that survive the robust estimation loop (i.e. RANSAC), or those supplied with the submission if using custom matches. As the dataset does not contain depth, we color them based on the symmetric epopilar distance on normalized coordinates: from 0 (green) to 2e-4 (yellow), and in red above it. Note that these occasionally contain false positives.
— Amsterdam —
— Bangkok —
— Barcelona —
— Buenos Aires —
— Cambridge —
— Cannes —
— Chicago —
— Helsinki —
— Madrid —
— Mountain View —
— New Orleans —
— San Francisco —
— Singapore —
— Sydney —
— Tokyo —
— Toronto —
— Zurich —
Google Urban dataset / Multiview track
mAA at 10 degrees: N/A (±0.00000 over 1 run(s) / ±0.10558 over 17 scenes)
Rank (per category): 49 (of 87)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| AMS | 2048.0 | 154.10 | 77.61 Rank: 45/87 |
1054.67 Rank: 39/87 |
3.620 Rank: 41/87 |
19.85048 Rank: 7/87 |
0.20207 (±0.00000) Rank: 46/87 |
0.26446 (±0.00000) Rank: 46/87 |
| BGK | 2048.0 | 205.43 | 92.80 Rank: 44/87 |
1886.95 Rank: 34/87 |
3.779 Rank: 5/87 |
11.37594 Rank: 7/87 |
0.02054 (±0.00000) Rank: 46/87 |
0.06697 (±0.00000) Rank: 45/87 |
| BCN | 2048.0 | 92.32 | 71.28 Rank: 45/87 |
545.72 Rank: 45/87 |
3.390 Rank: 46/87 |
22.70495 Rank: 9/87 |
0.06327 (±0.00000) Rank: 48/87 |
0.10280 (±0.00000) Rank: 48/87 |
| BAR | 2048.0 | 109.25 | 65.05 Rank: 45/87 |
527.12 Rank: 44/87 |
3.083 Rank: 45/87 |
11.44794 Rank: 8/87 |
0.06232 (±0.00000) Rank: 46/87 |
0.11325 (±0.00000) Rank: 45/87 |
| CAM | 2048.0 | 127.21 | 74.80 Rank: 44/87 |
930.58 Rank: 35/87 |
3.356 Rank: 42/87 |
22.28860 Rank: 13/87 |
0.08795 (±0.00000) Rank: 49/87 |
0.13595 (±0.00000) Rank: 48/87 |
| CAN | 2048.0 | 170.42 | 68.93 Rank: 39/87 |
676.18 Rank: 33/87 |
3.372 Rank: 37/87 |
18.94606 Rank: 25/87 |
0.02026 (±0.00000) Rank: 47/87 |
0.05171 (±0.00000) Rank: 45/87 |
| CHI | 2048.0 | 53.30 | 60.77 Rank: 49/87 |
311.57 Rank: 39/87 |
3.374 Rank: 49/87 |
41.90665 Rank: 8/87 |
0.01225 (±0.00000) Rank: 49/87 |
0.03156 (±0.00000) Rank: 49/87 |
| HEL | 2048.0 | 279.46 | 85.59 Rank: 49/87 |
1826.36 Rank: 36/87 |
3.647 Rank: 47/87 |
9.12742 Rank: 1/87 |
0.20830 (±0.00000) Rank: 45/87 |
0.32237 (±0.00000) Rank: 45/87 |
| MAD | 2048.0 | 82.36 | 74.75 Rank: 47/87 |
614.58 Rank: 44/87 |
3.355 Rank: 8/87 |
3.58906 Rank: 6/87 |
0.02034 (±0.00000) Rank: 48/87 |
0.05430 (±0.00000) Rank: 48/87 |
| MTV | 2048.0 | 90.37 | 68.62 Rank: 50/87 |
599.93 Rank: 46/87 |
3.414 Rank: 47/87 |
10.59311 Rank: 5/87 |
0.06527 (±0.00000) Rank: 49/87 |
0.11379 (±0.00000) Rank: 50/87 |
| NOR | 2048.0 | 93.30 | 55.74 Rank: 47/87 |
433.53 Rank: 44/87 |
3.000 Rank: 46/87 |
7.70669 Rank: 9/87 |
0.06302 (±0.00000) Rank: 45/87 |
0.09337 (±0.00000) Rank: 45/87 |
| SF | 2048.0 | 121.16 | 68.36 Rank: 49/87 |
765.92 Rank: 44/87 |
3.212 Rank: 47/87 |
9.71791 Rank: 4/87 |
0.07823 (±0.00000) Rank: 49/87 |
0.13075 (±0.00000) Rank: 49/87 |
| SG | 2048.0 | 126.40 | 58.40 Rank: 48/87 |
604.37 Rank: 39/87 |
3.306 Rank: 28/87 |
9.48295 Rank: 4/87 |
0.02831 (±0.00000) Rank: 49/87 |
0.05544 (±0.00000) Rank: 49/87 |
| SYD | 2048.0 | 95.06 | 69.09 Rank: 47/87 |
428.06 Rank: 48/87 |
3.298 Rank: 40/87 |
9.55832 Rank: 7/87 |
0.08816 (±0.00000) Rank: 47/87 |
0.13309 (±0.00000) Rank: 47/87 |
| TOK | 2048.0 | 148.23 | 80.04 Rank: 47/87 |
1175.88 Rank: 44/87 |
3.700 Rank: 31/87 |
10.99726 Rank: 8/87 |
0.34800 (±0.00000) Rank: 47/87 |
0.41948 (±0.00000) Rank: 47/87 |
| TOR | 2048.0 | 56.32 | 58.36 Rank: 49/87 |
300.91 Rank: 48/87 |
3.016 Rank: 46/87 |
7.49853 Rank: 4/87 |
0.02441 (±0.00000) Rank: 48/87 |
0.05062 (±0.00000) Rank: 48/87 |
| ZRH | 2048.0 | 103.44 | 61.28 Rank: 49/87 |
689.58 Rank: 49/87 |
3.207 Rank: 45/87 |
10.64905 Rank: 4/87 |
0.00859 (±0.00000) Rank: 50/87 |
0.02676 (±0.00000) Rank: 50/87 |
| Avg | 2048.0 | 124.01 | 70.09 Rank: 49/87 |
786.58 Rank: 43/87 |
3.361 Rank: 46/87 |
13.96711 Rank: 6/87 |
0.08243 (±0.00000) Rank: 49/87 |
0.12745 (±0.00000) Rank: 49/87 |
In the multi-view track we reconstruct the scene with Structure-from-Motion (Colmap) with small sets of images. We show the results for one bag of 10 images. Keypoints are drawn in blue if they are part of the model, and in red otherwise.
— Amsterdam —
— Bangkok —
— Barcelona —
— Buenos Aires —
— Cambridge —
— Cannes —
— Chicago —
— Helsinki —
— Madrid —
— Mountain View —
— New Orleans —
— San Francisco —
— Singapore —
— Sydney —
— Tokyo —
— Toronto —
— Zurich —
